Bump IREE to iree-org/iree@402a9be46a

[Abhishek-Varma]

-- This commit bumps IREE to iree-org/iree@402a9be46a. -- As part of the same it :-
a. removes "dangling" builders
b. adds explicit cast
c. adds explicit result type of enum

Signed-off-by: Abhishek Varma [email protected]

[Abhishek-Varma]

For build to succeed mlir-air needs to be updated.

Locally I have tried the following patch and the build succeeded :-

diff --git a/mlir/lib/Conversion/AIRLoweringPass.cpp b/mlir/lib/Conversion/AIRLoweringPass.cpp
index 3922980..b3d5ae6 100644
--- a/mlir/lib/Conversion/AIRLoweringPass.cpp
+++ b/mlir/lib/Conversion/AIRLoweringPass.cpp
@@ -1182,7 +1182,11 @@ public:
       // Label root of perfectly nested affine for loops for affine
       // optimizations.
       std::vector<SmallVector<mlir::affine::AffineForOp, 6>> bands;
-      mlir::affine::getTileableBands(f, &bands);
+      for (mlir::affine::AffineForOp forOp : f.getOps<mlir::affine::AffineForOp>()) {
+        SmallVector<mlir::affine::AffineForOp, 6> band;
+        getPerfectlyNestedLoops(band, forOp);
+        bands.push_back(band);
+      }
       for (auto &band : bands) {
         band[0]->setAttr("affine_opt_label",
                          StringAttr::get(f.getContext(), "tiling"));

CC: @erwei-xilinx could you please take a look ? I tried raising a PR for mlir-air but the bump seemed involved hence sharing the patch above which worked locally.

[erwei-xilinx]

CC: @erwei-xilinx could you please take a look ? I tried raising a PR for mlir-air but the bump seemed involved hence sharing the patch above which worked locally.

Sure, let me take a look.

[erwei-xilinx]

@Abhishek-Varma mlir-air has updated its llvm version: Xilinx/mlir-air#1056